Characterising processing conditions that artifactually bias human brain tissue transcriptomes

Area of Science:

• Neuroscience
• Genomics
• Bioinformatics

Background:

• Autopsy brain tissue is vital for neurobiology research.
• Post-mortem handling introduces transcriptional artifacts, complicating data interpretation.
• Standardized methods are needed to account for tissue degradation over time and temperature variations.

Purpose of the Study:

• To characterize transcriptomic changes in human brain tissue based on post-mortem intervals (PMIs).
• To identify and define artifact-responsive genes, termed Brain Artifact Genes (BAGs).
• To develop a predictive signature (TTRUTH) for quantifying processing effects in brain RNA-seq data.

Main Methods:

• Compared transcriptomic signatures of immediately extracted brain tissue with short (~6h) and long (~36h) PMIs.
• Mapped artifact gene signatures onto single-nucleus RNA-seq data.
• Utilized deep learning to create the TTRUTH predictive signature.

Main Results:

• Significant gene signature deviations (BAGs) were observed in both short and long PMIs compared to immediate extraction.
• Glutamatergic neurons showed early induction of artifact genes, followed by oligodendrocytes.
• The TTRUTH signature effectively quantifies processing-related transcriptional heterogeneity.

Conclusions:

• Post-mortem time and temperature significantly impact brain transcriptomic data.
• The TTRUTH signature provides a tool to standardize and interpret autopsy-derived brain RNA-seq datasets.
• This work enhances data reliability for neurobiological research and sample stratification.